Skip to content

Enable rocm/7.2.1 #67

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
michaelmckinsey1 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Enable rocm/7.2.1 #67

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,7 +63,7 @@ cuda = [
"mpi4py==4.1.1",
]
rocm = [
"torch==2.10.0+rocm7.1",
"torch==2.10.0+rocm7.2.1.lw.gitb07cec22",
"mpi4py==4.1.1",
]
rocmwci = [
Expand Down
31 changes: 0 additions & 31 deletions scripts/install-rccl.sh
View file
Open in desktop

This file was deleted.

9 changes: 5 additions & 4 deletions scripts/install-tuolumne-torchpypi.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
. install-rccl.sh
ml load python/3.11.5 && python3 -m venv .venvs/scaffoldvenv-tuo-pypi && source .venvs/scaffoldvenv-tuo-pypi/bin/activate && pip install --upgrade pip
ml cce/21.0.0 cray-mpich/9.1.0 rocm/7.1.1 rccl/fast-env-slows-mpi
pip install -e .[rocm] --prefix=.venvs/scaffoldvenv-tuo-pypi --extra-index-url https://download.pytorch.org/whl/rocm7.1 2>&1 | tee install.log
ml load python/3.12.2 && python3 -m venv .venvs/scaffoldvenv-tuo-pypi && source .venvs/scaffoldvenv-tuo-pypi/bin/activate && pip install --upgrade pip
ml cce/21.0.0 cray-mpich/9.1.0 rocm/7.2.1 rccl/fast-env-slows-mpi
pip install -e .[rocm] --find-links https://repo.radeon.com/rocm/manylinux/rocm-rel-7.2.1/ 2>&1 | tee install.log
# libmpi.so.12 does not exist => ls /opt/cray/pe/lib64/ | grep libmpi
patchelf --replace-needed libmpi.so.12 libmpi_gnu.so.12 .venvs/scaffoldvenv-tuo-pypi/lib/python3.12/site-packages/mpi4py/MPI.mpich.cpython-312-x86_64-linux-gnu.so
2 changes: 1 addition & 1 deletion scripts/install-tuolumne.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
ml load python/3.11.5 && python3 -m venv .venvs/scaffoldvenv-tuo && source .venvs/scaffoldvenv-tuo/bin/activate && pip install --upgrade pip
ml cce/21.0.0 cray-mpich/9.1.0 rocm/7.1.1 rccl/fast-env-slows-mpi
pip install -e .[rocmwci] --prefix=.venvs/scaffoldvenv-tuo 2>&1 | tee install.log
pip install -e .[rocmwci] 2>&1 | tee install.log
# Needed until new wheel exists for torch using mpich 9.1.0
TORCH_LIB_DIR=".venvs/scaffoldvenv-tuo/lib/python3.11/site-packages/torch/lib"
OLD="libmpi_gnu_112.so.12"
Expand Down
11 changes: 5 additions & 6 deletions scripts/scaffold-tuolumne-torchpypi.job
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,21 +5,20 @@
# flux: -g=1
# flux: -t 60m
# flux: -qpdebug
# flux: -B fractale
# flux: -B flask

ml cce/21.0.0 cray-mpich/9.1.0 rocm/7.1.1 rccl/fast-env-slows-mpi
ml cce/21.0.0 cray-mpich/9.1.0 rocm/7.2.1 rccl/fast-env-slows-mpi

. .venvs/scaffoldvenv-tuo-pypi/bin/activate

# Use ccl plugin that we manually built with install-rccl.sh
export NCCL_NET_PLUGIN=../aws-ofi-nccl.git/install/lib/librccl-net.so
export NCCL_NET_PLUGIN=/collab/usr/global/tools/rccl/toss_4_x86_64_ib_cray/rocm-7.2.0/install/lib/librccl-net.so

# Disable direct convolution benchmarking (should speedup warmup by a significant amount, does the below three options together)
# export MIOPEN_DEBUG_CONV_DIRECT=0
# Disable direct naive convolution benchmarking (naive_conv_ab_nonpacked_fwd_ndhwc_half_double_half.kd)
export MIOPEN_DEBUG_CONV_DIRECT_NAIVE_CONV_FWD=0
# Disable naive_conv_ab_nonpacked_bwd_ndhwc_half_double_half.kd
# export MIOPEN_DEBUG_CONV_DIRECT_NAIVE_CONV_BWD=0
export MIOPEN_DEBUG_CONV_DIRECT_NAIVE_CONV_BWD=0
# Disable naive_conv_ab_nonpacked_wrw_ndhwc_half_double_half.kd
export MIOPEN_DEBUG_CONV_DIRECT_NAIVE_CONV_WRW=0

Expand All @@ -29,4 +28,4 @@ torchrun-hpc -N 1 -n 1 $(which scaffold) generate_fractals -c $(pwd)/ScaFFold/co
#export PROFILE_TORCH=ON

torchrun-hpc -N 1 -n 4 --gpus-per-proc 1 $(which scaffold) benchmark -c $(pwd)/ScaFFold/configs/benchmark_default.yml
#torchrun-hpc -N 2 -n 4 --gpus-per-proc 1 $(which scaffold) benchmark -c $(pwd)/ScaFFold/configs/benchmark_default.yml
# torchrun-hpc -N 2 -n 4 --gpus-per-proc 1 $(which scaffold) benchmark -c $(pwd)/ScaFFold/configs/benchmark_default.yml
6 changes: 3 additions & 3 deletions scripts/scaffold-tuolumne.job
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@
# flux: -g=1
# flux: -t 60m
# flux: -qpdebug
# flux: -B fractale
# flux: -B flask

ml cce/21.0.0 cray-mpich/9.1.0 rocm/7.1.1 rccl/fast-env-slows-mpi

Expand All @@ -20,7 +20,7 @@ export LD_PRELOAD="/opt/rocm-7.1.1/llvm/lib/libomp.so /opt/cray/pe/mpich/9.1.0/o
# Disable direct naive convolution benchmarking (naive_conv_ab_nonpacked_fwd_ndhwc_half_double_half.kd)
export MIOPEN_DEBUG_CONV_DIRECT_NAIVE_CONV_FWD=0
# Disable naive_conv_ab_nonpacked_bwd_ndhwc_half_double_half.kd
# export MIOPEN_DEBUG_CONV_DIRECT_NAIVE_CONV_BWD=0
export MIOPEN_DEBUG_CONV_DIRECT_NAIVE_CONV_BWD=0
# Disable naive_conv_ab_nonpacked_wrw_ndhwc_half_double_half.kd
export MIOPEN_DEBUG_CONV_DIRECT_NAIVE_CONV_WRW=0

Expand All @@ -30,4 +30,4 @@ torchrun-hpc -N 1 -n 1 $(which scaffold) generate_fractals -c $(pwd)/ScaFFold/co
#export PROFILE_TORCH=ON

torchrun-hpc -N 1 -n 4 --gpus-per-proc 1 $(which scaffold) benchmark -c $(pwd)/ScaFFold/configs/benchmark_default.yml
#torchrun-hpc -N 2 -n 4 --gpus-per-proc 1 $(which scaffold) benchmark -c $(pwd)/ScaFFold/configs/benchmark_default.yml
# torchrun-hpc -N 2 -n 4 --gpus-per-proc 1 $(which scaffold) benchmark -c $(pwd)/ScaFFold/configs/benchmark_default.yml
Loading